knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = FALSE)

library(tidyverse)
library(rlang)
library(viridis)

theme_set(theme_minimal() + theme(legend.position = "right"))

options(
  ggplot2.continuous.colour = "viridis",
  ggplot2.continuous.fill = "viridis"
)

scale_colour_discrete = scale_colour_viridis_d
scale_fill_discrete = scale_fill_viridis_d
# Load and wrangle the overall data.

diabetes_df = read_csv("data/cleaned_diabetes_data.csv")

diabetes_df = 
  diabetes_df |> 
  mutate(
    has_diabetes = case_match(
      has_diabetes,
      0 ~ "Not diabetic",
      1 ~ "Pre-diabetic",
      2 ~ "Diabetic",
      3 ~ "Diabetic while pregnant",
      NA ~ NA),
    sex_at_birth = 
      case_match(
        sex_at_birth,
        1 ~ "male", 
        0 ~ "female"),
    sex_at_birth = fct_infreq(sex_at_birth),
    age_category = factor(age_category, levels = c("18-24", "25-29", "30-34", "35-39", "40-44", "45-49", "50-54", "55-59", "60-64", "65-69", "70-74", "75-79", "80+"), order = TRUE),
    type = case_when(
      sex_at_birth == "male" & !(age_category %in% c("18-24", "25-29")) & has_diabetes == "Diabetic" ~ 2,
      sex_at_birth == "female" & (
        (!(age_category %in% c("18-24", "25-29")) & pregnant == 0) | 
        !(age_category %in% c("18-24", "25-29", "30-34", "35-39", "40-44", "45-49"))
        ) & has_diabetes == "Diabetic" ~ 2,
      age_category %in% c("18-24", "25-29") & has_diabetes == "Diabetic" ~ 1,
      # If any of the conditions are not met (or NA), assign NA
      (is.na(sex_at_birth) | is.na(age_category) | is.na(pregnant)) & is.na(has_diabetes) ~ NA,
      TRUE ~ 0
    ),
    type = as.factor(type),
    diab_type = as.factor(diab_type),
    has_diabetes = as.factor(has_diabetes)
  )

Distribution of Different Diabetes Variables

There are three variables that we are interested in understanding – has_diabetes, diab_type, and type.

has_diabetes: This variable describes the people in the survey who responded after being asked the question, “Have you ever been told you had diabetes?” Overall, 432,339 people responded to this question.

diab_type: This variable comes directly from the dataset, where respondents who were diabetic per the has_diabetes question were asked “What type of diabetes do you have?” Only 22,027 participants answered this question out of the 59,786 people who responded that they had diabetes. Since there are over 60% of respondents unaccounted for, this measure may not tell us enough about risk factors or comorbidities of T2D in the general population.

type: Due to the missing data for diab_type, we created another variable type that describes the person’s diabetes type by using other demographic information from this dataset. We used this paper published by the CDC that used the same BRFSS survey from 2014 to classify T2D diagnosis. In this paper, they classified a survey respondent to have type II diabetes if the respondent was older than 30, not pregnant, and answered yes to the question “Have you ever been told you have diabetes?”

We then classified a respondent to have type II diabetes if the respondent was older than 30, not pregnant, and were diabetic as per the has_diabetes question. We classified a respondent to have type I diabetes if the respondent was younger than 30 and were diabetic as per the has_diabetes question.

Diabetes Status (has_diabetes)

diabetes_dist = function(df, var) {
  df |> 
    group_by({{ var }}) |> 
    summarize(count = n()) |> 
    knitr::kable()
}

diabetes_plot = function(df, var) {
  diabetes_df |> 
    ggplot(aes(x = {{ var }})) +
    geom_bar(aes(fill = factor({{ var }})), na.rm = FALSE) + 
    geom_text(
      stat = "count",  # Use the count statistic for frequencies
      aes(label = ..count..),  # Access the count directly
      vjust = -0.5,  # Adjust position of labels above the bars
      na.rm = FALSE
    ) + 
    labs(
      title = "Distribution of Diabetes", 
      x = "Diabetes Status", 
      y = "Count", 
      fill = as.character(rlang::ensym(var))) +
    theme_minimal() +
    theme(legend.position="bottom")
}

Table

diabetes_dist(df = diabetes_df, var = has_diabetes)
has_diabetes count
Diabetic 59786
Diabetic while pregnant 3253
Not diabetic 358706
Pre-diabetic 10594
NA 984

Graph

diabetes_plot(df = diabetes_df, var = has_diabetes)

Reported Type (diab_type)

Table

diabetes_dist(df = diabetes_df, var = diab_type)
diab_type count
1 1958
2 20069
NA 411296

Graph

diabetes_plot(df = diabetes_df, var = diab_type)

Evaluated Type (type)

Table

diabetes_dist(df = diabetes_df, var = type)
type count
0 372797
1 664
2 59007
NA 855

Graph

diabetes_plot(df = diabetes_df, var = type)

Distribution of Diabetes by Surveyed Demographics

diabetes_comorbidities = 
  diabetes_df |> 
  mutate(
    kidney_disease = case_match(kidney_disease, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    kidney_disease = as.factor(kidney_disease),
    heart_attack = case_match(heart_attack, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    heart_attack = as.factor(heart_attack),
    chd = case_match(chd, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    chd = as.factor(chd),
    stroke = case_match(stroke, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    stroke = as.factor(stroke),
    arthritis = case_match(arthritis, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    arthritis = as.factor(arthritis),
    asthma_ever = case_match(asthma_ever, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    asthma_ever = as.factor(asthma_ever),
    asthma_now = case_match(asthma_now, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    asthma_now = as.factor(asthma_now),
    covid_test = case_match(covid_test, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    covid_test = as.factor(covid_test),
    bronchitis = case_match(bronchitis, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    bronchitis = as.factor(bronchitis),
    high_bp = case_match(high_bp,
                         0 ~ "No", 1 ~ "Mild/Severe", 2 ~ "Yes, Severe", NA ~ NA),
    high_bp = as.factor(high_bp),
    high_bs = case_match(high_bs, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    high_bs = as.factor(high_bs),
    a1c_check = case_match(a1c_check, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    a1c_check = as.factor(a1c_check),
    high_chol = case_match(high_chol, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    high_chol = as.factor(high_chol),
    smoker = case_match(smoker, 0 ~ "No", 1 ~ "Yes", NA ~ NA),
    smoker = as.factor(smoker),
    type = as.factor(type),
    diab_type = as.factor(diab_type)
  ) |> 
  rename(
    `History of Kidney Disease` = kidney_disease,
    `History of Heart Attack` = heart_attack,
    `History of CHD` = chd,
    `History of Stroke` = stroke,
    `History of Arthritis` = arthritis,
    `History of Asthma` = asthma_ever,
    `Currently Have Asthma` = asthma_now,
    `Ever Had a Positive Covid Test` = covid_test,
    `History of Bronchitis` = bronchitis,
    `High Blood Pressure` = high_bp,
    `High Blood Sugar` = high_bs,
    `Checked for A1C in the Past Year` = a1c_check,
    `High Cholesterol` = high_chol,
    `Ever Been a Smoker` = smoker,
    `Has Diabetes` = has_diabetes,
    `Reported Type` = diab_type,
    `Evaluated Type` = type
  ) |> 
  select(`Has Diabetes`, `Reported Type`, `Evaluated Type`, everything())

Measures of Diabetes and Different Comorbidities

This section explores the distribution of comorbidities across diabetes diagnoses.

  • History of kidney disease
  • History of heart attack
  • History of coronary heart disease (CHD)
  • History of stroke
  • History of arthritis
  • History of asthma
  • Currently has asthma
  • Ever tested positive for COVID-19
  • History of bronchitis
# Define a function for the plot

comorbidities_plot = function(df, comorbidity, diabetes) {
  # Convert character strings to symbols
  comorbidity_sym <- rlang::sym(comorbidity)
  diabetes_sym <- rlang::sym(diabetes)
  
  diabetes_comorbidities |> 
    filter(!is.na(!!comorbidity_sym), !(is.na(!!diabetes_sym))) |> 
    group_by(!!comorbidity_sym, !!diabetes_sym) |> 
    summarize(n = n(), .groups = 'drop') %>%
    group_by(!!comorbidity_sym) %>%
    mutate(Percent = n / sum(n) * 100) |> 
    ggplot(aes(x = !!comorbidity_sym, y = Percent, fill = !!diabetes_sym)) +
    geom_bar(stat = "identity", position = "dodge") +
    geom_text(aes(label = sprintf("%.1f%%", Percent)),
              position = position_dodge(width = 0.9),
              size = 3.5, 
              vjust = -0.3) + 
    labs(
      title = str_c(diabetes, " by ", comorbidity), 
      x = comorbidity, 
      y = "Percent (%)", 
      fill = diabetes) +
    theme_minimal() +
    theme(legend.position="bottom")
}

Kidney disease

comorbidity = "History of Kidney Disease"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

## Something is wrong with diab_type!

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

## Convert to factor is being weird
comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

Heart Attack/MI

comorbidity = "History of Heart Attack"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

CHD

comorbidity = "History of CHD"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

Stroke

comorbidity = "History of Stroke"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

Arthritis

comorbidity = "History of Arthritis"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

Asthma

has_diabetes

comorbidities_plot(diabetes_comorbidities, "History of Asthma", "Has Diabetes")

comorbidities_plot(diabetes_comorbidities, "Currently Have Asthma", "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, "History of Asthma", "Reported Type")

comorbidities_plot(diabetes_comorbidities, "Currently Have Asthma", "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, "History of Asthma", "Evaluated Type")

comorbidities_plot(diabetes_comorbidities, "Currently Have Asthma", "Evaluated Type")

COVID-19

comorbidity = "Ever Had a Positive Covid Test"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")

Bronchitis

comorbidity = "History of Bronchitis"

has_diabetes

comorbidities_plot(diabetes_comorbidities, comorbidity, "Has Diabetes")

diab_type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Reported Type")

type

comorbidities_plot(diabetes_comorbidities, comorbidity, "Evaluated Type")